Round score
0.7263
+72.63%
Prompts scored
31 / 32
median_e2e_speedup
Below tolerance
1
answered too little to score
Zeroed total
1
scored 0.0 without measuring
The round score is the scoring rule applied to the prompts below. Both engines are compared at the same output token count, so a candidate that stops early cannot buy speed by answering less: a prompt under the tolerance bar scores nothing and shows its gate reason.
32 prompts
| Prompt | Speedup | Baseline | Candidate | Tokens | Gate |
|---|---|---|---|---|---|
| hf-000 | +57.58% | 0.591s | 0.251s | 34 | — |
| hf-001 | +73.82% | 0.713s | 0.187s | 41 | — |
| hf-002 | +61.21% | 0.794s | 0.308s |
32 requests
| Request | TTFT | Total | Tokens | Slowest gap | Stream |
|---|---|---|---|---|---|
| hf-000 | 115.66ms | 0.343s | 58 | 44.44ms | — |
| hf-001 | 82.16ms | 0.187s | 41 | 15.46ms | — |
| hf-002 | 126.49ms | 0.338s |
verdict
evidence
min logprob
num prompts
mean logprob
coverage ratio
quantile logprob
num positions scored
role
| 48 |
| — |
| hf-003 | +59.49% | 0.709s | 0.287s | 43 | — |
| hf-004 | +74.40% | 0.741s | 0.190s | 42 | — |
| hf-005 | +75.95% | 0.715s | 0.172s | 43 | — |
| hf-006 | +60.96% | 0.735s | 0.287s | 46 | — |
| hf-007 | +56.37% | 0.712s | 0.311s | 41 | — |
| hf-008 | — | — | — | 28 | candidate output below tolerance |
| hf-009 | +75.60% | 0.777s | 0.190s | 43 | — |
| hf-010 | +75.39% | 0.761s | 0.187s | 41 | — |
| hf-011 | +77.69% | 0.774s | 0.173s | 43 | — |
| hf-012 | +75.59% | 0.749s | 0.183s | 42 | — |
| hf-013 | +73.27% | 0.731s | 0.195s | 40 | — |
| hf-015 | +54.93% | 0.376s | 0.169s | 20 | — |
| hf-017 | +70.24% | 0.466s | 0.139s | 25 | — |
| hf-016 | +64.53% | 0.712s | 0.253s | 41 | — |
| hf-014 | +72.60% | 1.187s | 0.325s | 71 | — |
| hf-019 | +69.49% | 0.526s | 0.160s | 29 | — |
| hf-018 | +61.20% | 0.779s | 0.302s | 46 | — |
| hf-021 | +72.66% | 0.526s | 0.144s | 25 | — |
| hf-020 | +76.20% | 0.780s | 0.186s | 43 | — |
| hf-022 | +69.44% | 0.944s | 0.289s | 54 | — |
| hf-023 | +64.44% | 0.766s | 0.272s | 43 | — |
| hf-025 | +73.72% | 0.746s | 0.196s | 42 | — |
| hf-024 | +74.06% | 0.787s | 0.204s | 46 | — |
| hf-026 | +77.10% | 0.765s | 0.175s | 43 | — |
| hf-027 | +78.14% | 0.777s | 0.170s | 43 | — |
| hf-028 | +78.89% | 0.752s | 0.159s | 41 | — |
| hf-029 | +74.98% | 0.680s | 0.170s | 41 | — |
| hf-030 | +59.89% | 0.533s | 0.214s | 34 | — |
| hf-031 | +46.82% | 0.533s | 0.283s | 42 | — |
| 59 |
| 40.27ms |
| — |
| hf-003 | 91.98ms | 0.287s | 47 | 38.92ms | — |
| hf-004 | 87.43ms | 0.190s | 42 | 15.01ms | — |
| hf-005 | 67.58ms | 0.172s | 43 | 16.46ms | — |
| hf-006 | 105.14ms | 0.287s | 46 | 41.27ms | — |
| hf-007 | 86.99ms | 0.311s | 41 | 40.39ms | — |
| hf-008 | 83.49ms | 0.187s | 28 | 15.46ms | — |
| hf-009 | 70.97ms | 0.190s | 43 | 15.92ms | — |
| hf-010 | 69.02ms | 0.187s | 41 | 15.98ms | — |
| hf-011 | 54.92ms | 0.173s | 43 | 16.03ms | — |
| hf-012 | 64.77ms | 0.183s | 42 | 16.20ms | — |
| hf-013 | 76.52ms | 0.195s | 40 | 16.18ms | — |
| hf-014 | 59.43ms | 0.325s | 71 | 61.41ms | — |
| hf-015 | 108.84ms | 0.254s | 46 | 40.64ms | — |
| hf-016 | 108.94ms | 0.279s | 46 | 40.37ms | — |
| hf-017 | 79.92ms | 0.324s | 67 | 54.73ms | — |
| hf-018 | 108.49ms | 0.302s | 46 | 39.96ms | — |
| hf-019 | 86.89ms | 0.190s | 42 | 15.51ms | — |
| hf-020 | 67.03ms | 0.186s | 43 | 16.58ms | — |
| hf-021 | 54.87ms | 0.144s | 25 | 16.55ms | — |
| hf-022 | 63.89ms | 0.289s | 54 | 39.56ms | — |
| hf-023 | 88.51ms | 0.272s | 43 | 38.81ms | — |
| hf-024 | 72.59ms | 0.204s | 46 | 15.11ms | — |
| hf-025 | 77.56ms | 0.196s | 42 | 16.44ms | — |
| hf-026 | 71.55ms | 0.175s | 43 | 16.09ms | — |
| hf-027 | 65.14ms | 0.170s | 43 | 17.60ms | — |
| hf-028 | 55.09ms | 0.159s | 41 | 20.67ms | — |
| hf-029 | 64.53ms | 0.170s | 41 | 16.58ms | — |
| hf-030 | 109.10ms | 0.353s | 64 | 41.02ms | — |
| hf-031 | 93.08ms | 0.283s | 42 | 17.18ms | — |
evidence
requests per s
sla goodput ratio
output tokens per s
p99 e2e ms rel range
p99 itl ms rel range
p99 ttft ms rel range
Image digest
Patch hash